Every major AI chatbot — including ChatGPT, Claude, Gemini, Copilot, and Grok — uses consumer conversations as training material by default. Users can disable this in about a minute per service, and because personalization and memory are managed separately, opting out should not affect day-to-day use. The controls are usually buried in privacy settings, and most platforms label them with language that frames data collection positively.
Opting out has limits. It does not apply retroactively, so conversations already used for training remain in the models. It also does not delete stored data, which is governed by each platform's retention policy. Anthropic, for example, keeps conversations for up to five years when training is enabled, but drops that to 30 days after opt-out. Google retains chats reviewed by human reviewers for up to three years, even if a user deletes their history.
Meta stands apart. US users have no way to opt out of training on their Meta AI chats, and since December the company has also used those conversations for ad targeting across Facebook, Instagram, WhatsApp, and Messenger. Users in the UK and EU can object under GDPR, but for US users the only practical option is to avoid engaging with Meta AI entirely.